Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/xzawed/claude-grok-build-pluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/xzawed/claude-grok-build-plugin/delegate)<a href="https://agentmods.dev/commands/xzawed/claude-grok-build-plugin/delegate"><img src="https://agentmods.dev/badge/commands/xzawed/claude-grok-build-plugin/delegate/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/xzawed/claude-grok-build-plugin/delegate"><img src="https://agentmods.dev/badge/commands/xzawed/claude-grok-build-plugin/delegate.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00007 | $0.00900 |
| Opus 5.5 | $0.00003 | $0.00360 |
| Sonnet 5.5 | $0.00001 | $0.00180 |
| Haiku 4.5 | $0.00001 | $0.00090 |
Grade A, and why
delegate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Call grok_build_status first with the task's absolute cwd (or grok_auth_check with that
cwd when you only need auth) — grok resolves a relative GROK_HOME against the folder it runs
in, so the check looks there. (With worktree: true grok runs in a new worktree instead; the
delegation re-checks that folder itself and explains a refusal.) If ready is
false, stop and show the message. If it reports billingMismatch, tell the user that the
server is in subscription mode but their delegation history contains metered runs, so
GROK_BUILD_AUTH_MODE was api for some of them. The flag is about past history, not a
prediction that the next call will bill as metered.
If it reports billingCaveat, show its message before delegating: grok's config.toml
gives some model its own key or one it inherits from [model_providers.<id>] (or could not be
checked), so a run on that model may be billed to that key even though billing says
subscription. It is a warning — do not stop.
If the task's fit for Grok is unclear, call grok_build_route and follow nextAction.
When it says handle_with_claude, do not force Grok. When
nextAction.requiresHumanGateBeforeDelegate is set, run grok_build_plan and wait for the
user's approval before any edit. If the plan result has planWroteFiles or committed set to true,
the tree already changed before any approval — say so first.
Otherwise call grok_build_delegate with the user's task as prompt and an absolute cwd.
For wide or risky edits pass worktree: true, so changes land in an isolated worktree instead
of the working tree.
If status is not completed (auth_error, timeout, or grok_error), show the
returned message and stop — do not report the run as done; filesChanged may still
list partial edits.
Show the returned summary, filesChanged, and — importantly — the billing field, so the
user sees which mode the server was configured in. It is a tag derived from
GROK_BUILD_AUTH_MODE, not a measured charge. If the result carries billingCaveat, show
its message beside billing. If the result includes
sessionId, note that a follow-up can use /grok:resume (or resume on the next
delegate). A completed result can also carry a message — show it whenever there is one: besides
a commit notice, it can say that the prompt, with any files it @-mentions or skill it invokes, was too long for grok (its
request to the model left out part of the prompt or of what grok added to it), so the work
may not reflect the whole prompt.
If committed is true, grok made a git commit although this plugin never commits — show the
returned message: it names the folder whose HEAD moved and the commit it moved from, with how
to inspect and undo it there (never run those commands in another folder; in a worktree: true
run the commit is on the worktree's branch, which grok_build_worktree diff/apply do not carry).
Do not read an empty or short filesChanged as "nothing changed": committed files are no longer
listed. If committed is absent, the run could not be checked — do not report "no commit". If resumedCwd is
present (a resume or continue run), the session belongs to that directory and grok worked
THERE, not in the cwd you passed — review the diff there.
Finish at the review gate: run /grok:review on the diff. Do not commit — the user decides
accept, fix, or discard.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago Changed · +3 lines ac59a84607f1
- 17d ago Changed · +1 lines 9f1ff356db26
- 22d ago Changed · +1 lines 008d9b826316
- 28d ago Changed · +9 lines 376d4d0731d3
- 1mo ago Changed · +8 lines 5f29421f8451
- 1mo ago Changed · +4 lines 5be97a8274ab
- 1mo ago First seen · 28 lines · 7 tokens per session scan A bd72d2d4f03c
delegate is a command published in the GitHub repository xzawed/claude-grok-build-plugin (3 stars, last pushed today), licensed MIT. It adds 7 tokens to every session and 900 once invoked, about $0.0000 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
merge
Finish a PR properly: every check green, every review addressed — human and bot — then merge and clean up.
spec
Spec-first design: a gap-closing interview that produces a complete spec, with a quality controller that blocks until every section is answered and every question resolved.
debug
Systematic debugging: root cause before any fix, one hypothesis at a time, and a three-strikes rule that questions the architecture instead of stacking patches.
plan
Turn an approved spec into an implementation plan an engineer with zero context could execute — with a quality controller that blocks placeholders and hollow tasks.
tdd
Test-driven development with tests that actually catch breaks: red before green, name the break each test catches, and the mutation check before done.
todo
The quality-gated task list: tasks with real descriptions, testable acceptance criteria, and evidence — a task only closes when the controller agrees it is done.